← Meta Interview Insights

Meta·Data Scientist·Technical Phone Screen·Intermediate

Intermediate
May 2026

Summary

Stats fundamentals check for a Data Scientist role at Meta. Three questions, all centered on CLT and confidence intervals, with a whiteboard component at the end. Pretty textbook stuff but the whiteboard part tripped me up more than I expected.

Questions Asked (3)

Q1

Can you state the Central Limit Theorem and explain why it's relevant to A/B testing?

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

I knew the definition cold but fumbled connecting it to A/B testing in a crisp way.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly stating the Central Limit Theorem (CLT) in simple terms, emphasizing that the sampling distribution of the mean approaches normality as sample size grows, regardless of the population distribution. Then, connect it directly to A/B testing by explaining how CLT underpins the validity of hypothesis tests and confidence intervals used to compare metrics between control and treatment groups.

Pro tip: Mention that while CLT is powerful, it requires independent and identically distributed (i.i.d.) observations and finite variance; in practice, A/B tests often violate these assumptions (e.g., due to network effects or heavy-tailed metrics), so it's crucial to check sample size and consider robust alternatives.

1. State the CLT

Define the Central Limit Theorem: For a sufficiently large sample size, the distribution of the sample mean will be approximately normal, regardless of the underlying population distribution, provided observations are independent and identically distributed.

2. Explain relevance to A/B testing

Describe how A/B tests compare means (e.g., conversion rates, revenue per user) between two groups. The CLT allows us to assume that the difference in means follows a normal distribution, enabling the use of t-tests and z-tests to determine statistical significance.

3. Discuss practical implications

Highlight that CLT justifies the calculation of confidence intervals and p-values, which are essential for making data-driven decisions. It also informs sample size calculations to ensure sufficient power.

4. Address assumptions and limitations

Mention that CLT requires large enough sample sizes (often n ≥ 30 per group, but depends on skewness) and independence. In real-world A/B tests, violations like network effects or ratio metrics can occur, so alternative methods (e.g., bootstrap, delta method) may be needed.

Key Points to Mention

  • Definition of CLT: sampling distribution of the mean tends to normality as sample size increases.
  • A/B testing relies on comparing means between control and treatment groups.
  • CLT enables parametric tests (t-test, z-test) for hypothesis testing and confidence intervals.
  • Sample size requirements: larger samples lead to better normal approximation.
  • Assumptions: independence, identical distribution, finite variance.
  • Limitations: heavy-tailed metrics, network effects, and ratio metrics may require alternative approaches.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

Given a sample mean of 3, a known population standard deviation of 1, and n = 400, build a 95% confidence interval for the mean.

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

Plugged in 3 plus or minus 1.96 times 1 divided by 20 and got the interval pretty fast.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

State the formula for a confidence interval for a population mean with known standard deviation: x̄ ± z*(σ/√n). Plug in the given values (x̄=3, σ=1, n=400) and compute the standard error and margin of error. Then interpret the interval in the context of the problem, emphasizing what 95% confidence means.

Pro tip: Mention that with a large sample size (n=400), the Central Limit Theorem ensures the sampling distribution is approximately normal, so the z-interval is valid even if the population is not normal. Also, briefly discuss how this relates to A/B testing at Meta, where such intervals are used to estimate treatment effects.

1. Identify the given information

Extract the sample mean (x̄ = 3), population standard deviation (σ = 1), sample size (n = 400), and confidence level (95%).

2. Determine the critical value

For a 95% confidence level, the critical z-value is 1.96 (from the standard normal distribution).

3. Compute the standard error and margin of error

Standard error = σ/√n = 1/√400 = 0.05. Margin of error = z* × SE = 1.96 × 0.05 = 0.098.

4. Construct the confidence interval

The 95% confidence interval is x̄ ± margin of error = 3 ± 0.098, resulting in (2.902, 3.098).

5. Interpret the interval

We are 95% confident that the true population mean lies between 2.902 and 3.098. In an A/B testing context, this could represent the range of plausible values for the average treatment effect.

Key Points to Mention

  • Formula for confidence interval with known population standard deviation: x̄ ± z*(σ/√n).
  • Use of z-distribution (standard normal) because population standard deviation is known and sample size is large.
  • Calculation of standard error: σ/√n = 1/20 = 0.05.
  • Critical z-value for 95% confidence: 1.96.
  • Margin of error: 1.96 * 0.05 = 0.098.
  • Interpretation: We are 95% confident the true mean is between 2.902 and 3.098; this does not mean 95% of the data lie in this interval.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

Sketch the sampling distribution on the whiteboard and mark where the confidence interval bounds fall.

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

This was the awkward part.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly defining the sampling distribution (e.g., of the sample mean or difference in means) and its key properties (center, spread, shape). Then sketch the distribution on the whiteboard, marking the mean and the confidence interval bounds, and explain how the bounds relate to the standard error and critical value. Finally, connect the sketch to the context of A/B testing at Meta, emphasizing interpretation and practical implications.

Pro tip: Always label the axes and annotate the confidence level (e.g., 95%) and the margin of error on the sketch. This shows attention to detail and helps communicate the concept clearly to non-technical stakeholders.

1. Define the sampling distribution

State that the sampling distribution represents the distribution of a statistic (e.g., sample mean or difference in means) over repeated sampling. Mention its mean equals the population parameter and its standard deviation is the standard error.

2. Sketch the distribution

Draw a normal curve (if CLT applies) centered at the estimated parameter. Label the x-axis with the metric (e.g., conversion rate difference) and the y-axis as density or frequency.

3. Mark the confidence interval bounds

Indicate the lower and upper bounds of the confidence interval on the x-axis. Shade the middle area corresponding to the confidence level (e.g., 95%) and mark the critical values (e.g., ±1.96 SE) from the center.

4. Explain the relationship

Describe how the bounds are calculated: point estimate ± (critical value × standard error). Clarify that the interval captures the true parameter with the specified confidence over repeated samples.

5. Connect to A/B testing context

Relate the sketch to the experiment: the sampling distribution of the difference in means between control and treatment. Discuss how the CI helps decide if the effect is statistically significant (e.g., if it excludes zero).

Key Points to Mention

  • Central Limit Theorem and normality assumption for large samples
  • Standard error and its role in determining interval width
  • Confidence level (e.g., 95%) and critical value (z or t)
  • Interpretation: the interval contains the true parameter in 95% of repeated samples
  • Relationship to hypothesis testing: if CI excludes zero, effect is significant
  • Practical significance vs. statistical significance in product decisions

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.